A classification assigns invariants to objects so that two objects receive the same invariant exactly when they are equivalent. Requiring this assignment to be Borel measurable makes it possible to compare classification problems without choosing arbitrary representatives. Borel reducibility expresses this comparison: a relation reduces to a relation when a single Borel map preserves and reflects equivalence for every pair of inputs. Su Gao's Invariant Descriptive Set Theory, Section 15.2 and Figure 15.1, collects benchmark relations that separate several levels of classification complexity.
This mission concerns the comparisons asserted in Section 1 of Ruijun Wang's supplied manuscript Comparisons of benchmark equivalence relations. The manuscript is the source of the proposed new results. Its claims and the completeness of their Lean proofs must be distinguished: some earlier sequence-space results have accepted proofs, while the added countable-relation, jump, and operator comparisons remain formalization targets. The requested category is OpenProblem; here the outstanding tasks concern complete formal proofs of the stated manuscript results. Source and upload files are intended for the Benchmark repository.
Let have the product Borel structure. Its elements are arbitrary real sequences. For a finite real exponent , the relation identifies when is finite. The relation identifies them when , and identifies them when is finite. No boundedness or convergence condition is imposed on either input separately.
For a relation on a space , its countable product compares two sequences coordinate by coordinate. Its Friedman–Stanley jump compares the sets of -classes represented by their coordinates: every coordinate of either sequence must be -equivalent to some coordinate of the other sequence. Repetitions and ordering do not matter, and the witnessing matches need not form a bijection of indices.
The benchmark is presented by the full shift of the free group on . Thus means for some and every . The presentation includes points with nontrivial stabilizers. The relation is mutual absolute continuity of Borel probability measures on Cantor space , with the usual evaluation sigma-algebra on probabilities.
The operator and representation targets use the fixed complex Hilbert space . All bounded linear operators on are allowed. Representations are all unitary representations on this space, including reducible representations. Their Borel structures come from convergence on every vector, and, for representations, at every group element. The representation notation follows the manuscript; this superscript denotes arbitrary representations rather than the jump operation above.
The goal is the conjunction of the six main comparisons in the introduction. Each component is also a named milestone:
The symbol requires both nonreductions; requires a reduction in the forward direction and a nonreduction in the reverse direction. Significant established comparisons are retained as literature milestones, including the classical positive range and Thomas's nonreduction .
The comparisons describe which benchmark classifications admit invariants of another specified kind. The operator result, if fully formalized, rules out scalar measure classes as Borel complete invariants for arbitrary bounded operators. It makes no claim against other kinds of spectral analysis or operator-valued invariants. Formalization must establish the comparisons themselves, rather than infer them from a shared Borel complexity bound.
A nonreduction quantifies over every Borel map. Checking a coordinate formula, a continuous family, or a finite collection of inputs cannot establish it. The countable-relation result requires the descriptive and analytic infrastructure connecting its source action to sequence-space equivalence. The jump assertions require control of arbitrary countable sets of target classes. The operator comparisons additionally require measurability, exact preservation and reflection of equivalence, and transport to the fixed Hilbert space. Each of these requirements is part of the target, even when supporting theory is absent from Mathlib.
The common definitions use actual free groups, actual real sequences, real powers, convergence, probability measures, bounded complex-linear maps, and surjective complex-linear isometries. An infinite exponent is represented by the top element of WithTop applied to the reals; its relation is bounded difference. Measurable structures for operators and representations are specified explicitly to avoid accidental use of a stronger operator topology.
The target environment is Lean 4.30.0 with supported Mathlib revision c5ea00351c28e24afc9f0f84379aa41082b1188f. Exact definition and statement files are compiled locally before upload. A compiling open statement is an elaboration check, not a proof. Completed solutions must contain no proof placeholders or additional mathematical axioms. In particular, assuming a reduction, a rigidity conclusion, a local-row property, or a classification theorem that the argument is meant to establish does not complete the corresponding headline result.
The existing accepted theorem identities are reused where they already express the intended result. Reusable work includes Hellinger affinity, absolute continuity, probability kernels, compact quotient obstructions, sequence codings, unitary representations, countable-row arguments, and operator intertwiners. Further contributions should preserve the exact quantifiers, endpoint cases, and source and target Borel structures.
Benchmark.tex, Section 1, Theorems 1.1–1.6. Repository designated for sources.theorem BenchmarkComparisons.all_main_comparisons :
CountableComparisons ∧ ExponentThreshold ∧ NullMeasureObstruction ∧
JumpObstructions ∧ OperatorStrictComparison ∧ RepresentationComparisons := by sorry
The conjunction of Theorems 1.1–1.6: and for all finite ; the threshold exactly for , including ; ; both nonreductions to ; strict measure-to-operator reducibility; and the full chain involving arbitrary representations of and . All relations, Borel structures and quantifiers are those fixed by the shared definitions.
No open leaves. Every sub-goal is proved or awaiting decomposition.